CN110750756B - Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm - Google Patents

Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm Download PDF

Info

Publication number
CN110750756B
CN110750756B CN201910946059.2A CN201910946059A CN110750756B CN 110750756 B CN110750756 B CN 110750756B CN 201910946059 A CN201910946059 A CN 201910946059A CN 110750756 B CN110750756 B CN 110750756B
Authority
CN
China
Prior art keywords
fuzzy
equation
support vector
vector machine
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910946059.2A
Other languages
Chinese (zh)
Other versions
CN110750756A (en
Inventor
郝富强
陈珺逸
戴旺
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Xingjian Software Co ltd
Shenzhen Wellreach Automation Co ltd
Original Assignee
Shenzhen Xingjian Software Co ltd
Shenzhen Wellreach Automation Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Xingjian Software Co ltd, Shenzhen Wellreach Automation Co ltd filed Critical Shenzhen Xingjian Software Co ltd
Priority to CN201910946059.2A priority Critical patent/CN110750756B/en
Publication of CN110750756A publication Critical patent/CN110750756A/en
Application granted granted Critical
Publication of CN110750756B publication Critical patent/CN110750756B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D18/00—Testing or calibrating apparatus or arrangements provided for in groups G01D1/00 - G01D15/00
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Mathematical Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Evolutionary Biology (AREA)
  • Computational Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

An immediate on-line instrument checksum diagnosis method through an optimal support vector machine algorithm comprises the following steps: s1, constructing a flow net model; s2, iterating field actual measurement data, and calculating and determining parameters in the model through an optimal support vector machine algorithm to enable the model to be usable; s3, restarting the steps regularly, and optimizing parameters; s4, checking the sampled variables one by using the model in a stable flow field state; s5, after the suspected failure points are eliminated, performing inverse iterative operation by using the rest data, and reversely deducing theoretical calculation values of the suspected failure points; s6, eliminating process condition changes, comparing and analyzing actual instrument signals by using the theoretical calculation value, realizing verification and fault diagnosis, and determining signal health level; s7, recording sampling signals and calculation signals according to the measurement time, and alarming and positioning faults according to deterministic fault diagnosis conditions. The invention can realize early detection and early report of instrument faults, intelligently correct results and improve working efficiency.

Description

Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm
Technical Field
The invention relates to an instant on-line meter checksum diagnostic method.
Background
In recent years, the intellectualization and automation of industrial production are increasingly emphasized. In the intelligent manufacturing process, the intelligence of the meter is an important component. Currently mainstream meters adopt manual periodicity to detect one by one to judge, and the staff can't in time accurately judge whether the instrument measured value is accurate to the opportunity of handling has been musied, and then whole production activity is influenced. When the instrument works, intelligent diagnosis of the traditional instrument or the electronic equipment is only aimed at the instrument, and open loop self-verification can only be carried out, so that the accuracy of data and whether a streaming network system operates normally cannot be verified.
Disclosure of Invention
The invention aims to provide an instant on-line instrument checksum diagnosis method through an optimal support vector machine algorithm
The invention can realize the aim by designing an instant on-line instrument check sum diagnosis method through an optimal support vector machine algorithm, which comprises the following steps:
s1, constructing a flow network model comprising a flow channel model and an equipment assembly model through a hydrodynamic continuity equation, a momentum equation and an energy equation;
first the flow equation is reduced to
F＝(1-K 0 )*a 1 *(P 1 -P 2 -KZ)+K 0 *F 1p
wherein ,
Figure GDA0004194502020000011
wherein ,
Figure GDA0004194502020000012
is the pressure from the last iteration, kz=ρg (Z 2 -Z 1 ) Wherein ρ is fluid density, g is gravitational acceleration, Z 1 For elevation at point 1, Z 2 Is the elevation at point 2; f (F) 1p Value F obtained by the last iteration; k (K) 0 For a user selectable constant, K can be adjusted by 0 Obtaining stability of numerical solution;
in the above, F, P 1 and P2 For the unknown quantity, the height difference KZ is a system constant, and the rest is a value obtained by the last iteration and can be considered as the known quantity;
a mass balance equation is also set, wherein the inflow node is (+) and the outflow node is (-);
s2, iterating field actual measurement data, and calculating and determining parameters in the model through an optimal support vector machine algorithm to enable the model to be usable;
according to the matrix equation set formed in step S1, the pair F (F 3 ) Factors influencing the calculation of the value are taken as model input, and the F value is taken as output;
s3, restarting the steps regularly, and optimizing model parameters so as to adapt to new working conditions again, so that the model is automatically learned and maintained;
s4, utilizing the model obtained in the step, and checking the sampled variables one by one in a stable flow field state;
s5, after the suspected failure points are eliminated, performing inverse iterative operation by using the rest data, and reversely deducing theoretical calculation values of the suspected failure points;
s6, eliminating process condition changes, comparing and analyzing actual instrument signals by using the theoretical calculation value, obtaining deviation parameters of the actual signals by adopting a predefined fault mode and deviation evaluation, and realizing verification and fault diagnosis by threshold judgment, fuzzy logic and fault hypothesis verification to determine the signal health level;
s7, recording sampling signals and calculation signals according to the measurement time, and realizing alarming and fault positioning according to diagnosis conditions of a flow network knowledge base and an instrument fault feature base.
Further, determining membership of the fuzzy equation;
let the fuzzy equation system have c * The centers of the fuzzy groups k and j are v respectively k 、v j Then the ith training sample X i Membership μ for fuzzy group k ik The method comprises the following steps:
Figure GDA0004194502020000021
wherein n is a blocking matrix index required in the fuzzy classification process, and is usually taken as 2; II is a norm expression;
using the above membership value or its variants to obtain a new input matrix;
for the fuzzy group k, its input matrix is deformed as:
φ ik (X i ,μ ik )＝[1 func(μ ik )X i ]
wherein func (μ) ik ) For membership value mu ik Is generally taken as
Figure GDA0004194502020000031
φ ik (X i ,μ ik ) Representing the ith input variable X i And membership mu of its fuzzy group k ik A new input matrix corresponding to the input matrix;
taking the weighted least square support vector machine as a local equation of a fuzzy equation system, and carrying out optimization fitting on each fuzzy group; let the ith target output of the model training sample be F i The weighted support vector machine equates the fitting problem to the quadratic programming problem by transformation;
Figure GDA0004194502020000032
Figure GDA0004194502020000033
Figure GDA0004194502020000034
where R (ω, ζ) is the objective function of the optimization problem, minR (ω, ζ) is the minimum of the objective function of the optimization problem,
Figure GDA0004194502020000035
nonlinear mapping function, N is training sample number, ζ= { ζ 1 ,…,ξ N "is a relaxation variable, ζ i Is the ith component of the relaxation variable, ω is the normal vector to the support vector machine hyperplane, b is the corresponding offset, and ω i And gamma is the least squares support direction, respectivelyWeight and penalty factor of measuring machine, +.>
Figure GDA0004194502020000036
Is the i-th component ζ of the weighted least squares support vector machine relaxation variable i Estimating a standard deviation; c 1 Is a constant, here taken as 2.5; c 2 Is a constant, here taken as 3; the superscript T denotes transpose, μ ik Representing training sample X i Membership of fuzzy group k, phi ik (X i ,μ ik ) Representing the ith input variable X i And membership mu of its fuzzy group k ik A new input matrix corresponding to the input matrix;
from the above, the output of the fuzzy group k in the training sample i is derived as follows:
Figure GDA0004194502020000041
wherein
Figure GDA0004194502020000042
To blur group K at output of training sample i, K<·>Is a kernel function of a least squares support vector machine, where K<·>Taking a linear kernel function mu mk Representing the mth training sample X m Membership of fuzzy group k, phi mk (X m ,μ mk ) Represents the mth input variable X m And membership mu of its fuzzy group k mk Corresponding new input matrix alpha m Is the mth component of the corresponding lagrangian multiplier.
Further, a particle swarm algorithm is adopted to optimize a penalty factor and an error tolerance value of a local equation of a weighted least square support vector machine in a fuzzy equation, and the optimization steps are as follows:
s201, determining the optimization parameters of the particle numbers as penalty factors and error tolerance values of local equations of a weighted least square support vector machine, the individual number pop ize of the particle swarm, and the maximum cyclic optimization frequency item max Initial position r of the p-th particle p Initial velocity v p Local optimum value Lbest p And a global optimum Gbest for the entire population of particles;
s202, setting an optimization objective function, converting the optimization objective function into fitness, and evaluating each local fuzzy equation; calculating a fitness function through the corresponding error function, and considering that the particle fitness with large error is small, and the fitness function of the particle p is expressed as:
f p ＝1/(E p +1)
in the formula ,Ep Is an error function of the fuzzy equation,
Figure GDA0004194502020000043
in the formula ,
Figure GDA0004194502020000044
is the predictive output of the fuzzy equation system, F i Target output for the fuzzy equation system;
s203, circularly updating the speed and the position of each particle according to the following formula,
v p (iter+1)＝ω×v p (iter)+m 1 a 1 (Lbet p -r p (iter))+m 2 a 2 (Gbest-r p (iter))；
r p (iter+1)＝r p (iter)+v p (iter+1)；
in the formula ,vp Representing the velocity of the update particles p, r p Lbest represents the individual optimum value of the updated particle p, gbest represents the global optimum value of the whole particle swarm, iter represents the number of cycles, ω is the inertia weight in the particle swarm algorithm, m 1 、m 2 Corresponding acceleration coefficient, a 1 、a 2 Is [0,1 ]]Random numbers in between;
s204, for the particle p, if the new fitness is larger than the original individual optimal value, updating the individual optimal value of the particle: lbest p ＝f p ；
S205, if the individual optimum value Lbest of particle p p Is larger than the original granuleSub-group global optimum value Gbest, then gbest=lbest p ；
S206, judging whether the performance requirement is met, if yes, ending the optimizing to obtain a set of local equation parameters of the optimized fuzzy equation; otherwise, returning to the step S203, continuing the iterative optimization until the maximum iterative number item is reached max 。
Further, the periodicity in step S3 is defined as monthly or quarterly or annually.
Further, the variable in step S4 is a meter signal; recording measurement time, and comparing the calculated value with a measured value corresponding to the measurement time to obtain the percentage or variance or mean square error of the deviation range; after the complete verification is performed a plurality of times, the instrument is considered to be possibly invalid according to the deterministic fault diagnosis condition.
Further, theoretical calculation value P of suspected failure point i The formula of (c) is given by,
Figure GDA0004194502020000051
wherein ,Pi 、P j Indicating the pressure measured by the ith and jth sensors, Z i 、Z j Representing the elevation at the i and j th positions, F ij The mass flow rate between i and j is represented, ρ is represented by the fluid density, g is represented by the gravitational acceleration, and a is the flow coefficient.
Further, the predefined failure modes include drift, leakage, blockage, and failure modes; the flow network knowledge base comprises energy transfer characteristics of flow network nodes and branches; the instrument fault feature library comprises numerical drift, abnormal change rate, open circuit and short circuit fault features.
The invention combines the algorithm and the computer intelligent analysis, replaces the traditional manual inspection by month or quarter, can realize early detection, early report and intelligent correction of the faults of the instrument, greatly saves manpower and material resources and improves the working efficiency. Meanwhile, when partial meters are maintained offline due to faults, the invention can calculate the numerical value of an offline monitoring point by using the built flow network model and the readings of a sensor which normally works, and the normal operation of the system is not influenced.
Drawings
FIG. 1 is a flow chart of a preferred embodiment of the present invention;
FIG. 2 is a schematic illustration of a fluid network in accordance with a preferred embodiment of the present invention.
Detailed Description
The invention is further described below with reference to examples.
As shown in fig. 1, an on-line meter checksum diagnosis method in real time by an optimal support vector machine algorithm includes the following steps:
s1, constructing a flow network model comprising a flow channel model and an equipment assembly model through a hydrodynamic continuity equation, a momentum equation and an energy equation.
And constructing a flow network model by using a node method through a hydrodynamic continuity equation, a momentum equation (a Navier-Stokes equation) and an energy equation. For large-scale flow networks, simplifying the large-scale flow network or system into a plurality of small flow networks or systems can be adopted, so that the modeling flow is simplified.
In order to obtain an easy-to-calculate fluid network model, it is assumed that the fluid flows uniformly only along the catheter direction and that the response to changes in boundary conditions is very rapid. For compressible fluids, the node mass will increase or decrease depending on the actual operating conditions, assuming that the mass of the incoming conduit is not equal to the mass of the outgoing conduit. Compressibility and mass balance terms are introduced into the equation.
Figure GDA0004194502020000061
Wherein: f=mass flow rate=ρva, ρ=fluid density, v=flow rate, a=conduit cross-sectional area, x=conduit flow length, p=node pressure, t=node absolute temperature, α=compression coefficient.
The conservation of momentum equation can be written over the pipe length L:
Figure GDA0004194502020000062
wherein ：P1 ,P 2 Pressure at points 1,2, Z 1 ,Z 2 The elevation at points 1,2, ρ=fluid density, g=gravitational acceleration, H L Head loss of the length L of the pipe, v=flow rate,
the head loss term HL, i.e. the sum of all the main head losses due to friction effects and the small head losses due to inlet, fittings, area variations etc., can be expressed generally as a square of the fluid
Proportion: ρ gHL =f 2 /a 2 (3)
Wherein: a is calculated from the fluid flow rate, pressure drop and height difference.
Substituting (3) into (2)
Figure GDA0004194502020000071
Using a quasi-stable simplification, the last term is omitted and the equation is simplified to
Figure GDA0004194502020000072
The flow equation can be expressed as
F＝a[P 1 -P 2 -KZ] 1/2 (6)
Wherein: kz=ρg (Z 2 -Z 1 ) (7)
Equation (6) defines the relationship between the conduit flow rate and pressure.
A fluid network such as that shown in fig. 2 may be assumed to be a collection of closed pipes. Writing an equation as in equation (6) for each flow term results in a series of second order equations. To obtain pressure and flow in the network, these equations as well as the node mass balance equations must be solved simultaneously. For this purpose, the second order equation must first be linearized.
Formula (6) can be linearized into
F＝a 1 *[P 1 -P 2 -KZ] (8)
wherein
Figure GDA0004194502020000073
wherein
Figure GDA0004194502020000074
Is the pressure from the last iteration
Attempting to numerically solve a set of simultaneous equations such as equation (8) sometimes results in non-convergence of the iterative results. In order to guide the stability of the numerical solution, it is necessary to be in range
Figure GDA0004194502020000075
The relaxation factor Ko is introduced and equation (8) is modified as follows:
F＝a 1 *(P 1 -P 2 -KZ)-K 0 [a 1 *(P 1 -P 2 -KZ)-F 1p ] (9)
wherein ：
F 1p value F obtained by last iteration
Simplifying the above process to obtain
F＝(1-K 0 )*a 1 *(P 1 -P 2 -KZ)+K 0 *F 1p (10)
In practical application, K 0 Becomes a user selectable constant by adjusting K 0 And obtaining stability of numerical solution. K reduction 0 Physically can be considered as introducing inertia in the system.
In formula (10), F, P 1 and P2 Is an unknown quantity. The height difference KZ is a system constant, and the remainder is a value obtained by the previous iteration and can be regarded as a known quantity. KZ is typically ignored for simplicity of calculation.
As with the flow net in fig. 2, equation (10) can be expressed as follows:
Figure GDA0004194502020000082
Figure GDA0004194502020000083
Figure GDA0004194502020000084
Figure GDA0004194502020000085
Figure GDA0004194502020000086
in addition to momentum balance, mass balance equations are also required. Also, for the example problem in fig. 2, it is possible to give:
F 1 +F 2 -F 3 ＝0(16)
F 3 -F 4 -F 5 ＝0(17)
in the above formula, the inflow node is (+) and the outflow node is (-).
Equations (11) through (17) provide a complete set of seven equations for seven unknown independent variables, F 1 ，F 2 ，F 3 ，F 4 ，F 5 ，P 1 and P2 . In this problem, it is assumed that the boundary pressure P is given B Is known. The system of equations in matrix form is shown below.
Figure GDA0004194502020000081
All F lps The last iteration delivers a value that is considered to be known in the current time step.
S2, iterating the field actual measurement data, and calculating and determining a parameter F in the model through an optimal support vector machine algorithm to enable the model to be usable. The calculation process is as follows:
according to the matrix equation set above,will be opposite F (F 3 ) Various factors (P 1 、P 2 、P B 、P C 、P D 、P E Six modeling variables) as model inputs and F values as outputs.
Establishing this blur model includes the following 3 parts:
(1) And (3) determining membership of a fuzzy equation: let the fuzzy equation system have c * The centers of the fuzzy groups k and j are v respectively k 、v j Then the ith training sample X i Membership μ for fuzzy group k ik The method comprises the following steps:
Figure GDA0004194502020000091
where n is the index of the blocking matrix required in the fuzzy classification process, it is generally taken as 2 and, II is a norm expression.
Using the above membership value or its variants to obtain a new input matrix, for the fuzzy group k, its input matrix is deformed as:
φ ik (X i ,μ ik )＝[1 func(μ ik )X i ] (18)
wherein func (mu) ik ) For membership value mu ik Is generally taken as
Figure GDA0004194502020000092
Wait for phi ik (X i ,μ ik ) Representing the ith input variable X i And membership mu of its fuzzy group k ik The corresponding new input matrix.
(2) And taking the weighted least square support vector machine as a local equation of the fuzzy equation, and performing optimization fitting on each fuzzy group. Let the ith target output of the model training sample be F i The weighted support vector machine equates the fitting problem to the following quadratic programming problem by transforming:
Figure GDA0004194502020000093
Figure GDA0004194502020000094
Figure GDA0004194502020000095
where R (ω, ζ) is the objective function of the optimization problem, minR (ω, ζ) is the minimum of the objective function of the optimization problem,
Figure GDA0004194502020000096
nonlinear mapping function, N is training sample number, ζ= { ζ 1 ,…,ξ N "is a relaxation variable, ζ i Is the ith component of the relaxation variable, ω is the normal vector to the support vector machine hyperplane, b is the corresponding offset, and ω i And gamma is the weight and penalty factor of the least squares support vector machine, respectively,/>
Figure GDA0004194502020000097
Is the i-th component ζ of the weighted least squares support vector machine relaxation variable i Estimating a standard deviation; c 1 Is a constant, here taken as 2.5; c 2 Is a constant, here taken as 3; the superscript T denotes transpose, μ ik Representing training sample X i Membership of fuzzy group k, phi ik (X i ,μ ik ) Representing the ith input variable X i And membership mu of its fuzzy group k ik The corresponding new input matrix.
From (19) (20) (21), the output of the fuzzy group k in the training sample i can be deduced as follows:
Figure GDA0004194502020000101
wherein
Figure GDA0004194502020000102
To blur group K at output of training sample i, K<·>Is a kernel function of a least squares support vector machine, where K<·>Taking a linear kernel function mu mk Representing the mth training sample X m Membership of fuzzy group k, phi mk (X m ,μ mk ) Represents the mth input variable X m And membership mu of its fuzzy group k mk Corresponding new input matrix alpha m Is the mth component of the corresponding lagrangian multiplier.
(3) The particle swarm optimization module is used for optimizing the penalty factors and the error tolerance values of the least square support vector machine local equation in the fuzzy equation by adopting a particle swarm algorithm, and comprises the following specific implementation steps:
s201, determining that the optimization parameters of the particle numbers are penalty factors and error tolerance values of a least square support vector machine local equation, the individual number pop ize of the particle swarm, and the maximum cyclic optimization frequency item max Initial position r of the p-th particle p Initial velocity v p Local optimum value Lbest p And a global optimum Gbest for the entire population of particles.
S202, setting an optimization objective function, converting the optimization objective function into fitness, and evaluating each local fuzzy equation; calculating a fitness function through the corresponding error function, and considering that the particle fitness with large error is small, and the fitness function of the particle p is expressed as:
f p ＝1/(E p +1) (23)
in the formula ,Ep Is an error function of the fuzzy equation, expressed as:
Figure GDA0004194502020000103
in the formula ,
Figure GDA0004194502020000104
is the predictive output of the fuzzy equation system, F i Target output for the fuzzy equation system;
s203, circularly updating the speed and the position of each particle according to the following formula,
v p (iter+10＝ω×v p (iter)+m 1 a 1 (Lbest p -r p (iter00+m 2 a 2 (Gbest-r p (iter)) (25)
r p (iter+1)＝r p (iter)+v p (iter+1) (26)
in the formula ,vp Representing the velocity of the update particles p, r p Lbest represents the individual optimum value of the updated particle p, gbest represents the global optimum value of the whole particle swarm, iter represents the number of cycles, ω is the inertia weight in the particle swarm algorithm, m 1 、m 2 Corresponding acceleration coefficient, a 1 、a 2 Is [0,1 ]]Random numbers in between;
s204, for the particle p, if the new fitness is larger than the original individual optimal value, updating the individual optimal value of the particle:
Lbest p ＝f p (27)
s205, if the individual optimum value Lbest of particle p p Is greater than the original global optimum value Gbest of the particle swarm:
Gbest＝Lbest p (28)
s206, judging whether the performance requirement is met, if yes, ending the optimizing to obtain a set of local equation parameters of the optimized fuzzy equation; otherwise, returning to the step S203, continuing the iterative optimization until the maximum iterative number item is reached max 。
S3, restarting the model in the step S1-S2 periodically (monthly/quarterly/annual), and optimizing the model parameters so as to adapt to new working conditions again, and enabling the model to learn and maintain autonomously.
S4, utilizing the model obtained in the step, and checking the sampled variables (measuring instrument signals) one by one in a stable flow field state. Recording the measurement time, and comparing the calculated value with the measured value corresponding to the measurement time to obtain the percentage (or variance, mean square error, etc.) of the deviation range. After the complete verification is performed a plurality of times, the instrument is considered to be possibly invalid according to the deterministic fault diagnosis condition.
S5, after the suspected failure point is eliminated, performing inverse iterative operation by using the rest data, and reversely deducing a theoretical calculation value of the suspected failure point.
From equation (5):
Figure GDA0004194502020000111
wherein Pi 、P j Indicating the pressure measured by the ith and jth sensors, Z i 、Z j Representing the elevation at the i and j th positions, F ij Representing the mass flow rate between i, j.
S6, eliminating process condition changes, comparing and analyzing actual instrument signals by using the theoretical calculation value, obtaining deviation parameters of the actual signals by adopting a predefined fault mode and deviation evaluation, and realizing verification and fault diagnosis by threshold judgment, fuzzy logic and fault hypothesis verification to determine the signal health level; the predefined failure modes include drift, leakage, blockage, failure, etc. failure modes.
S7, recording sampling signals and calculation signals according to the measurement time, and realizing alarming and fault positioning according to diagnosis conditions of a flow network knowledge base and an instrument fault feature base. The flow network knowledge base includes energy transfer characteristics of the flow network nodes and branches. The instrument fault feature library comprises fault features such as numerical drift, abnormal change rate, open circuit, short circuit and the like.
The invention combines the algorithm and the computer intelligent analysis, replaces the traditional manual inspection by month or quarter, can realize early detection, early report and intelligent correction of the faults of the instrument, greatly saves manpower and material resources and improves the working efficiency. Meanwhile, when partial meters are maintained offline due to faults, the invention can calculate the numerical value of an offline monitoring point by using the built flow network model and the readings of a sensor which normally works, and the normal operation of the system is not influenced.

Claims (7)

1. The method for checking and diagnosing the real-time online instrument through the optimal support vector machine algorithm is characterized by comprising the following steps of:
s1, constructing a flow network model comprising a flow channel model and an equipment assembly model through a hydrodynamic continuity equation, a momentum equation and an energy equation;
first the flow equation is reduced to
F＝(1-K 0 )*a 1 *(P 1 -P 2 -KZ)+K 0 *F 1p
Wherein the linearization coefficient
Figure FDA0004194502010000011
wherein ,
Figure FDA0004194502010000012
is the pressure from the last iteration, kz=ρg (Z 2 -Z 1 ) Wherein ρ is fluid density, g is gravitational acceleration, Z 1 For elevation at point 1, Z 2 Is the elevation at point 2; f (F) 1p Value F obtained by the last iteration; k (K) 0 For a user selectable constant, K can be adjusted by 0 Obtaining stability of numerical solution;
in the above, F, P 1 and P2 For the unknown quantity, the height difference KZ is a system constant, and the rest is a value obtained by the last iteration and can be considered as the known quantity;
a mass balance equation is also set, wherein the inflow node is (+) and the outflow node is (-);
s2, iterating field actual measurement data, and calculating and determining parameters in the model through an optimal support vector machine algorithm to enable the model to be usable;
according to the matrix equation set formed in step S1, the pair F (F 3 ) Six factors P influencing the calculation of the value 1 、P 2 、P B 、P C 、P D 、P E Six modeling variables are used as model inputs, and F values are used as outputs;
s3, restarting the steps regularly, and optimizing model parameters so as to adapt to new working conditions again, so that the model is automatically learned and maintained;
s4, utilizing the model obtained in the step, and checking the sampled variables one by one in a stable flow field state;
s5, after the suspected failure points are eliminated, performing inverse iterative operation by using the rest data, and reversely deducing theoretical calculation values of the suspected failure points;
s6, eliminating process condition changes, comparing and analyzing actual instrument signals by using the theoretical calculation value, obtaining deviation parameters of the actual signals by adopting a predefined fault mode and deviation evaluation, and realizing verification and fault diagnosis by threshold judgment, fuzzy logic and fault hypothesis verification to determine the signal health level;
s7, recording sampling signals and calculation signals according to the measurement time, and realizing alarming and fault positioning according to diagnosis conditions of a flow network knowledge base and an instrument fault feature base.
2. The method for on-line meter checksum diagnosis on-line by optimal support vector machine algorithm according to claim 1, wherein: determining membership of a fuzzy equation;
let the fuzzy equation system have c * The centers of the fuzzy groups k and j are v respectively k 、v j Then the ith training sample X i Membership μ for fuzzy group k ik The method comprises the following steps:
Figure FDA0004194502010000021
wherein n is a blocking matrix index required in the fuzzy classification process, and is usually taken as 2; the terms are normative expressions;
using the above membership value or its variants to obtain a new input matrix;
for the fuzzy group k, its input matrix is deformed as:
φ ik (X i ，μ ik )＝[1func(μ ik )X i ]
wherein func (μ) ik ) For membership value mu ik Is generally taken as
Figure FDA0004194502010000022
φ ik (X i ，μ ik ) Representing the ith input variable X i And membership mu of its fuzzy group k ik A new input matrix corresponding to the input matrix;
taking the weighted least square support vector machine as a local equation of a fuzzy equation system, and carrying out optimization fitting on each fuzzy group; let the ith target output of the model training sample be F i The weighted support vector machine equates the fitting problem to the quadratic programming problem by transformation;
Figure FDA0004194502010000023
Figure FDA0004194502010000024
Figure FDA0004194502010000031
where R (ω, ζ) is the objective function of the optimization problem, minR (ω, ζ) is the minimum of the objective function of the optimization problem,
Figure FDA0004194502010000032
nonlinear mapping function, N is training sample number, ζ= { ζ 1 ，...，ξ N "is a relaxation variable, ζ i Is the ith component of the relaxation variable, ω is the normal vector to the support vector machine hyperplane, b is the corresponding offset, and ω i And gamma is the weight and penalty factor of the least squares support vector machine, respectively,/>
Figure FDA0004194502010000035
Is the i-th component ζ of the weighted least squares support vector machine relaxation variable i Estimating a standard deviation; c 1 Is a constant, here taken as 2.5; c 2 Is a constant, here taken as 3; the superscript T denotes transpose, μ ik Representing training sample X i Membership of fuzzy group k, phi ik (X i ，μ ik ) Representing the ith input variable X i And membership mu of its fuzzy group k ik A new input matrix corresponding to the input matrix;
from the above, the output of the fuzzy group k in the training sample i is derived as follows:
Figure FDA0004194502010000033
wherein
Figure FDA0004194502010000034
To blur group K at output of training sample i, K<·>Is a kernel function of a least squares support vector machine, where K<·>Taking a linear kernel function mu mk Representing the mth training sample X m Membership of fuzzy group k, phi mk (X m ，μ mmk ) Represents the mth input variable X m And membership mu of its fuzzy group k mmk Corresponding new input matrix alpha m Is the mth component of the corresponding lagrangian multiplier.
3. The method for checking and diagnosing the immediate online instrument by the optimal support vector machine algorithm according to claim 2, wherein the particle swarm algorithm is adopted to optimize the penalty factors and the error tolerance values of the weighted least squares support vector machine local equation in the fuzzy equation, and the optimization steps are as follows:
s201, determining the optimization parameters of the particle numbers as penalty factors and error tolerance values of local equations of a weighted least square support vector machine, the individual number pop ize of the particle swarm, and the maximum cyclic optimization frequency item max Initial position r of the p-th particle p Initial velocity v p Local optimum value Lbest p And a global optimum Gbest for the entire population of particles;
s202, setting an optimization objective function, converting the optimization objective function into fitness, and evaluating each local fuzzy equation; calculating a fitness function through the corresponding error function, and considering that the particle fitness with large error is small, and the fitness function of the particle p is expressed as:
f p ＝1/(E p +1)
in the formula ,Ep Is an error function of the fuzzy equation,
Figure FDA0004194502010000041
in the formula ,
Figure FDA0004194502010000042
is the predictive output of the fuzzy equation system, F i Target output for the fuzzy equation system;
s203, circularly updating the speed and the position of each particle according to the following formula,
v p (iter+1)＝ω×v p (iter)+m 1 a 1 (Lbest p -r p (iter))+m 2 a 2 (Gbest-r p (iter))；
r p (iter+1)＝r p (iter)+v p (iter+1)；
in the formula ,vp Representing the velocity of the update particles p, r p Lbest represents the individual optimum value of the updated particle p, gbest represents the global optimum value of the whole particle swarm, iter represents the number of cycles, ω is the inertia weight in the particle swarm algorithm, m 1 、m 2 Corresponding acceleration coefficient, a 1 、a 2 Is [0,1 ]]Random numbers in between;
s204, for the particle p, if the new fitness is larger than the original individual optimal value, updating the individual optimal value of the particle: lbest p ＝f p ；
S205, if the individual optimum value Lbest of particle p p If the particle swarm is larger than the original global optimal value Gbest, gbest=Lbest p ；
S206, judging whether the performance requirement is met, if yes, ending the optimizing to obtain a set of local equation parameters of the optimized fuzzy equation; otherwise, returning to the step S203, continuing the iterative optimization until the maximum iterative number item is reached max 。
4. The method for on-line meter checksum diagnosis on-line by optimal support vector machine algorithm according to claim 1, wherein: the period in step S3 is defined as monthly or quarterly or annually.
5. The method for on-line meter checksum diagnosis on-line by optimal support vector machine algorithm according to claim 1, wherein: the variable in the step S4 is a measuring instrument signal; recording measurement time, and comparing the calculated value with a measured value corresponding to the measurement time to obtain the percentage or variance or mean square error of the deviation range; after the complete verification is performed a plurality of times, the instrument is considered to be possibly invalid according to the deterministic fault diagnosis condition.
6. The method for on-line meter checksum diagnosis on-line by optimal support vector machine algorithm according to claim 1, wherein: theoretical calculated value of suspected failure point P i The formula of (c) is given by,
Figure FDA0004194502010000051
wherein ,Pi 、P j Indicating the pressure measured by the ith and jth sensors, Z i 、Z j Representing the elevation at the i and j th positions, F ij The mass flow rate between i and j is represented, ρ is represented by the fluid density, g is represented by the gravitational acceleration, and a is the flow coefficient.
7. The method for on-line meter checksum diagnosis on-line by optimal support vector machine algorithm according to claim 1, wherein: predefined failure modes include drift, leakage, blockage, and failure modes; the flow network knowledge base comprises energy transfer characteristics of flow network nodes and branches; the instrument fault feature library comprises numerical drift, abnormal change rate, open circuit and short circuit fault features.
CN201910946059.2A 2019-10-01 2019-10-01 Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm Active CN110750756B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910946059.2A CN110750756B (en) 2019-10-01 2019-10-01 Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910946059.2A CN110750756B (en) 2019-10-01 2019-10-01 Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm

Publications (2)

Publication Number Publication Date
CN110750756A CN110750756A (en) 2020-02-04
CN110750756B true CN110750756B (en) 2023-06-20

Family

ID=69277637

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910946059.2A Active CN110750756B (en) 2019-10-01 2019-10-01 Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm

Country Status (1)

Country Link
CN (1) CN110750756B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113778003A (en) * 2021-10-08 2021-12-10 沈阳仪表科学研究院有限公司 Debugging method of intelligent operation and maintenance platform data acquisition unit

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103675011A (en) * 2013-09-22 2014-03-26 浙江大学 Soft industrial melt index measurement instrument and method of optimal support vector machine
WO2018072351A1 (en) * 2016-10-20 2018-04-26 北京工业大学 Method for optimizing support vector machine on basis of particle swarm optimization algorithm
CN108764305A (en) * 2018-05-14 2018-11-06 浙江大学 A kind of improved colony intelligence machine learning fault diagnosis system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103675011A (en) * 2013-09-22 2014-03-26 浙江大学 Soft industrial melt index measurement instrument and method of optimal support vector machine
WO2018072351A1 (en) * 2016-10-20 2018-04-26 北京工业大学 Method for optimizing support vector machine on basis of particle swarm optimization algorithm
CN108764305A (en) * 2018-05-14 2018-11-06 浙江大学 A kind of improved colony intelligence machine learning fault diagnosis system

Also Published As

Publication number Publication date
CN110750756A (en) 2020-02-04

Similar Documents

Publication Publication Date Title
CN110441065B (en) Gas turbine on-line detection method and device based on LSTM
CN111258297B (en) Equipment health index construction and service life prediction method based on data fusion network
CN110197049B (en) Non-metal pipeline leakage positioning method based on transient inverse problem
CN107725283A (en) A kind of fan trouble detection method based on depth belief network model
CN105444923A (en) Mechanical temperature instrument error prediction method based on genetic-algorithm optimized least square support vector machine
KR100867938B1 (en) Prediction method for watching performance of power plant measuring instrument by dependent variable similarity and kernel feedback
CN113551744A (en) Ultrasonic flowmeter performance online monitoring method and system
CN112036075A (en) Abnormal data judgment method based on environmental monitoring data association relation
CN111860839A (en) Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm
CN115344019A (en) Natural gas metering flow adjusting process based on composite intelligent algorithm
CN110750756B (en) Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm
Jung et al. Systematic exploration of pipeline network calibration using transients
CN110705187B (en) Instant on-line instrument checksum diagnosis method through least square algorithm
CN110705186B (en) Real-time online instrument checksum diagnosis method through RBF particle swarm optimization algorithm
CN113988210A (en) Method and device for restoring distorted data of structure monitoring sensor network and storage medium
CN112096693B (en) On-line diagnosis and verification method for underwater production hydraulic control system, storage medium and control terminal
CN117034808A (en) Natural gas pipe network pressure estimation method based on graph attention network
Guan et al. Pressure-driven background leakage models and their application for leak localization using a multi-population genetic algorithm
CN117216703A (en) Water delivery pipe network operation data anomaly detection and correction method
CN117272202A (en) Dam deformation abnormal value identification method and system
CN117090831A (en) Hydraulic system fault diagnosis framework with twinning application layer
CN110750884B (en) Instant on-line instrument checksum diagnosis method through self-adaptive support vector machine algorithm
CN114462336B (en) Method for calculating average temperature of coolant of main pipeline of nuclear reactor
WO2019147986A1 (en) Determination of virtual process parameters
Yang et al. Monitoring and uncertainty analysis of feedwater flow rate using data-based modeling methods

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant